Predicting Graduate-level Performance from Undergraduate Achievements

One of the principal concerns in graduate admissions is how future performance in graduate studies may be predicted from a candidate’s undergraduate achievements. In this study, we examined the statistical relationship between B.Sc. and M.Sc. achievements using a dataset that is not subject to an admission-induced selection bias. Our analysis yielded three insights. First, we were able to explain 55% of the variance in M.Sc. performance using a small set of highly discriminative B.Sc. achievements alone. Second, we found the final B.Sc. year to be most informative for future M.Sc. performance. Third, when a failed exam was repeated, the crucial piece of information was the grade achieved in the final attempt. We envisage that results such as the ones described in this study may increasingly improve the design of future admission procedures.